Chapter 2
Theory
This chapter reviews the theoretical concepts of this work: First, the Hamiltonian
describing a molecule coupled with an external electric field is introduced in sec-
tion 2.1. In section 2.2 the adiabatic and diabatic representation of the time de-
pendent Schrodinger equation are introduced. The topics of the following sections
are rotational averaging (section 2.3), time dependent calculation of spectra (sec-
tion 2.4), pump-probe ionization spectroscopy (section 2.5), and the calculation of
the initial nuclear wave function (section 2.6). In section 2.7 the applied propaga-
tion schemes are described and section 2.8 deals with methods for finding an ap-
proximate solution of the electronic Schrodinger equation. Finally, our approach
of calculating the kinetic couplings is presented in section 2.9.
2.1 Schrodinger equation, molecular Hamiltonian
and coupling with the laser field
Time dependent Schrodinger equationThe time evolution of a state vector
���������is governed by the Schrodinger equa-
tion, � ���� � ����������� �� �����������������(2.1)
where�� �����
is the Hamilton operator of the system. In this work the Hamilton
operator consists of two parts: The first part describes the molecule without any
external field and neglecting relativistic effects. The second part expresses the
influence of the electric field (laser pulses in this work).
Molecular Hamilton operatorAssuming no interaction with the environment the non-relativistic time-
independent Hamilton operator of a molecule,��������
, with � nuclei and 19
Theory 20
electrons is given by!"�#�$�%'&)( !*,+,-/.0( !12+,-/.0(,!354 -/.�(6!784 -:9<; !=>&)( !*,+,-/.0(,!354 -:9@? !AB&C( !1D+,-/.�(6!784 -:9E. (2.2)
where !=>&)( !*�+F-/.0(,!3G4 -:9<;IH JK+5LNMPO QRTS + !* Q+ H UK V L 4 O QR:WYX !3 Q4 (2.3)
describes the kinetic energy of the nuclei and electrons and!AZ&)( !1D+F-/.0(6!784 -:9<; [\:]_^a` b H UK 4 LNM JK+5LNM c +6d Qe !784 H !1D+ e ? UK 4gfih d Qe !784 HI!78h e ? JK+ fkj c + c j d Qe !1D+ZH !1 j eml(2.4)
accounts for the electron-nuclei, electron-electron and nuclei-nuclei interactions,
respectively, and represents the potential energy of the molecule in the absence of
external time-dependent fields. In the above equations!1D+
is the position operator,!*�+the momentum operator,
S +the mass and c + the atomic number of the nuc-
leus n . Likewise!784 and
!3G4 are the operators of the position and momentum of the
electron o andWYX
is the electron mass.
Coupling with the Laser Field
The applied laser pulses used in this work can be described classically for they
have a high photon density. The electric field is a function of time p and space1
and can be written asq &r1 . p 9�;sd t q ` tau/& p 9vt d Vmwyx 1 z|{:}�~ ?�d z Vmw�x 1 z|{a}�~R .(2.5)
whered
denotes the polarization vector, q ` is the amplitude,u/& p 9 defines the shape
of the pulse,
xis the wave vector and � is the central frequency of the field. The
termd V�x 1
can be expanded in a Taylor series:d Vyx 1 ; [ ?�� x 1 ?��8�8�@�(2.6)
Since the spatial extension of a typical molecule is of the order of � 10 A whereas
the wave lengths of the applied lasers are around 400 nm (= 4000 A) the Taylor
expansion (2.6) can be truncated after the first term turning the electric field inde-
pendent of
xand
1: q & p 9�;�d t q ` t�u/& p 9�t d V {a} ?�d z V {a}R �
(2.7)
or: q & p 9<; q ` tau/& p 9vt��E��u/& ��p 9E. where q ` ;�d t q ` � (2.8)
Theory 21
It can be shown that, if the wave length of the electromagnetic field is much
smaller than the extension of the molecule, the light-molecule interaction con-
sists of an electric dipole, a magnetic-dipole and an electric-quadrupole part [50].
The latter two are by the order of the dimensionless factor 1/137 smaller than the
electric-dipole term [50]. This means that the electric-dipole term - if it does not
vanish at all - is the dominant one and the other two can be neglected. In this so-
called electric-dipole approximation [50, 51] the operator, ��������, describing the
interaction of the molecule with the electric field is given by��������<�I�2� �����g�_�� � �� � (2.9)
where �� is the electric dipole operator and��
is the rotation matrix that connects the
space-fixed coordinate system of� �����
with the molecule-fixed coordinate system
of �� . In this work it will be approximated by classical averaging (see section 2.3).
To describe the molecule-laser interaction the total Hamiltonian �� is written
as a sum of the molecular Hamiltonian ����v��� and the coupling with the external
electric field �� �����: ��¡�����<� ��������/¢ �� �����E£
(2.10)
Describing the electric field classically (cf equation (2.9)) and the molecule
quantum-mechanically (cf equation 2.2) is often referred to as the semiclassical
dipole approximation.
2.2 Adiabatic and diabatic representation of the
time-dependent Schrodinger equation
For a numerical treatment of the Mn-CO photodissociation the time-dependent
Schrodinger equation will be represented in the adiabatic basis in section 2.2.1 and
2.2.2. In section 2.2.3 the time-dependent equations of an alternative representa-
tion, the diabatic picture, will be derived. The probability of dissociation can be
used to determine life times of excited states. It will be defined in section 2.2.6
after the Born-Oppenheimer approximation has been introduced in section 2.2.5.
2.2.1 Derivation of the adiabatic representation
It is assumed that the eigenvectors ¤�¥�¦0§ and eigenvalues ¨ª© �¦ of the electronic
Schrodinger equation �� © � ¤�¥�¦E§ � ¨ © �¦ ¤�¥�¦0§ � (2.11)
Theory 22
where«¬®'¯±°³² ´µ ¶·N¸ ¹|º»:¼ «½ º¶�¾ ¿À�ÁFÂ�ÃÅÄ ² ´µ Æ ·5¶ÈÇµÉ ·N¸ Ê É6Ë ºÌ «Í ¶ ² «Î É Ì ¾ ´µ ¶gÏiÐ Ë ºÌ «Í ¶ ² «Í Ð ÌÒÑ (2.12)
is the electronic Hamiltonian, are known for fixed positions of the nuclei Ó °¿TÔÕÔ�Ô×Ö. In reality, exact solutions of (2.11) exist only for very simple systems.
In section 2.8 methods for an approximate solution of the electronic Schrodinger
equation are discussed.
Since the repulsion between nuclei,¿À�ÁFÂÃ Ä ÇµÉ ÏkØ Ê É Ê Ø Ë ºÌ «Î É ² «Î Ø Ì ÑÚÙis a constant for fixed nuclear coordinates it can be added to the electronic Hamilto-
nian «¬ÜÛ'¯ ° «¬®�¯ ¾ ¿À�Á@ÂÃÜÄ ÇµÉ ÏkØ Ê É Ê Ø Ë ºÌ «Î É ² «Î Ø Ì ÑÚÙ (2.13)
or to the electronic energy yielding the adiabatic potential energy ÝiÞ :Ý/Þ�ßCà «Î ÉFá:â °�ã '¯Þ ßCà «Î ÉFá:â ¾ ¿À�ÁFÂ�Ã Ä ÇµÉ ÏkØ Ê É Ê Ø Ë ºÌ «Î É ² «Î Ø Ì Ñ Ô(2.14)
With these definitions the molecular Hamiltonian reads as«¬�ä�å�¯±°I² ÇµÉ ·N¸ ¹/º»Tæ É «ç ºÉ ¾ «¬ Û'¯ Ô(2.15)
In the adiabatic representation the total wavefunctionÌ�è ß�é âê is expanded in
the basis of the electronic wavefunctionsÌ�ëíì ê which diagonalize the electronic
Hamiltonian
«¬®�¯:´0îkï à Î É á Ì�è ß�é âê ° µ Þñð Þ:ß)à Î É á Ù é â Ì�ë Þaß)à Î É á:âê �¯ Ô (2.16)
The multiplication ofÌ�è ß�é â�ê with ´0îiï à Î É á Ì on the left side of equation (2.16) de-
scribes a projection of the total wavefunction on the basis of the nuclear coordin-
ates. The index ” Ëóò ” in (2.16) denotes the dependence of the electronic wavefunc-
tionsÌ�ë Þ ê on the electronic coordinates. This index will be skipped in the follow-
ing equations. In equation (2.16) - the so called Born-Oppenheimer ansatz - the
time-dependent coefficients ð Þ�ß�é â can be interpreted as the nuclear wave functions.
Equation (2.16) is exact as long as the electronic basis à Ì�ë Þ:ß)à Î É á:â�êEá is not trun-
cated. Inserting (2.16) in the time dependent Schrodinger equation (2.1) with the
total Hamiltonian (2.10) leads to:ô ¹>õõ é µ Þñð Þ�ßCà Î É á Ù é â Ì�ë ÞaßCà Î É á:â�ê °
Theory 23
ö ÷øùûúNüþý|ÿ��� ù � ù ø � � ����� ù � ��������� ������ ù ����������� "! ��� ù ��� ø � � ����� ù � ��������� ����� ù �����
ö$# � ���&%(') %*�+ ��� ù ��� ø � � ����� ù � �����,�-� ����� ù �����/.Multiplying by 0 �21 ���� ù ����� on the left and integrating over the electronic coordin-
ates, one obtains 3 ý 44 � ø � � ������ ù � ����� 0 �21 ���� ù �����-� ������ ù �����5 687 9:<; = >ø � ?@A 0 �21 ��� ù ���,�-� ����� ù �����5 687 9:B; = ÷øùûúNü2C ö ý ÿ��� ù � ù � ����� ù � �����ED$F8GH
�JI � % ø � ÷øù5úNü ö ý/ÿI ��� ù ?@@A 0 �21 ��� ù ������K�ùL�-� ����� ù �����5 687 9MONQPSR; =UT VXWZY\[ V ]\^ %ZK�ù � ������ ù � ����� F8GGH� ø � ?@@A � ������ ù � �����_% ÷øù5úNü ?@@A ö ý/ÿ��� ù 0 �21 ��� ù ���,� � ù`��� ����� ù �����5 687 9M NbacR; =UT VXWdY\[ V ]\^ F GGH F GGH
� ø � ?@A 0 �21 ��� ù ����� ���� U! ��� ù ������� ������ ù �����5 687 9e = WZY\[ V ]\^gf :B; = % � ����� ù � �����F8GHö$# � ���/% ') % ø � ?@@A 0 �21 ���� ù ���,���+ ��� ù ������� ������ ù �����5 687 9h ; = WZY\[ V ]\^ % � ����� ù � ����� F8GGH �
where 0 �21 ���� ù �����-K�ùi�-� ����� ù ����� (2.17)
Theory 24
and j\k2lEm�n�oqp r�s,t�uvpit-kxw�mn�oqp r�s�y(2.18)
are the matrix elements of the kinetic coupling matrices z {S|U} and z {�~} , respectively.� lcorresponds to the adiabatic potential of the � -th electronic state, forming a di-
agonal element of the adiabatic potential matrix�
that is diagonal in the adiabatic
picture. Finally, � l-w mn�oqp r�s are the matrix elements of the adiabatic dipole matrix� � . With these notations the resulting adiabatic time dependent Schrodinger equa-
tion becomes:�E��������lmn�o$p r � � si���_� ��p�� | � ~��� p uvp�� � lm�n�oqp r�s� �
lmn�o$p r � � s� � w � ��pX� |� � ~� p z {�|U}l-w� p m�n�oqp r�s�����p�� �
w�mn�oqp r � � s �� � w � ��p�� |� � ~��� p z {Q~}l-wE� p m�n�oqp r�s � �
�w�m�n�oqp r � � s�q� m � s/�(�� � � w � l-w mn�oqp r�s_� �
w�mn�oqp r � � s/ (2.19)
2.2.2 Adiabatic description of the Mn-CO ¡8¢ photodissociation
Taking into account the Mn-CO £E¤ stretching coordinate ¥8£ with the corresponding
reduced mass �¦£ , �¦£ � §©¨�ª/«q¬ { ¨® }S¯ � §©¨®§ ¨�ª/«q¬ { ¨® }�¯ � §©¨® �(2.20)
and neglecting all other nuclear degrees of freedom, equation (2.19) gets
�� ������l�m ¥8£ � � sL� °±±±² � � ~� �¦£ � ~� ¥ ~£
� � lm ¥³£ s´ µ8¶ ·¸O¹ {Qº<»¼}½8¾¾¾¿ �
lm ¥8£ � � s´ µ8¶ ·Born-Oppenheimer dynamics (section 2.2.5)� � ~�¦£ � w z {�|U}l-w m ¥8£ s_� �� ¥8£
��w�m ¥8£ � � s_� � ~� �¦£ � w z {�~}l-w m ¥³£ s_� �
w�m ¥³£ � � s´ µ8¶ ·kinetic coupling terms
Theory 25À$Á Â<Ã�Ä/ÅÇÆÈ Å/É�ÊÌË Í Ê Â<Î8Ï8Ä_Å,Ð Ê Â<Î8Ï�Ñ�Ã�ÄÒ Ó8Ô Õcoupling by electric field
Ñ (2.21)
where ÖØ×�ÙUÚÍ Ê ÂBÎ8Ï8ÄiÛÌÜ\Ý Í ÂBγÏÞÄ,ßáàà Î8Ï ß-Ý Ê ÂBγÏÞÄ�â (2.22)
and Öã×QäÚÍ Ê Â<Î8Ï8ÄiÛÌÜ\Ý Í ÂBÎ8Ï8Ä�ß à äà Î äÏ ß�Ý Ê Â<Î8Ï8Ä�â³å (2.23)
The last equation describes the nuclear motion of a (pseudo-)diatomic molecule
along the internuclear bonding coordinate Î8Ï , where Ð Ê Â<Î8Ï�Ñ�Ã�ÄãÛæÐ(ç Í�èÊ ÂBÎ8Ï�Ñ�Ã�Ä is the
vibrational part of the nuclear wave function. The rotational and translational mo-
lecular degrees of freedom are separated by a product ansatz,Ð2Â<Î8Ï Ñ¼éëê ÄLÛìÐ ç ÍZè ÂBÎ8ϳÄ_Å,Ð(í�î"ï�ÂBðÞÏ�ѼñòϳÄ_Å,ÐóïZí ÏEôÞõ ÂUéqê Ä/Ñ (2.24)
where Î8Ï is the relative and éëê is the center of mass coordinate.
It is helpful to write equation (2.21) in matrix notation,öE÷ àà à РÛ�À ÷ äø ˦Ïúù à äà Î äÏ Ð ûýü Ð À ÷ äË¦Ï Ö ×�ÙUÚ àà Î³Ï Ð À ÷ äø Ë¦Ï Ö ×�äÚ Ð ÀþÁ ÂcÃ�Ä/ŦÆÈ Å�ÿ Ë Ð å(2.25)
In the adiabatic representation the potential matrix ü is diagonal with the potential
energy curves ü Í ÂBÎ8Ï³Ä of theö-th electronic state as diagonal elements:ü Û ��� ü Ù Â<Î8Ï8Ä � åÞåÞå� ü ä Â<Î8Ï8Ä
...
���� å (2.26)Ö ×�ÙUÚand
Ö ×�äÚdefine the off-diagonal part of the kinetic energy operator in the adia-
batic representation. In section 2.9 it is shown that
Ö ×�ÙUÚis antisymmetric with di-
agonal elements equal to zero,Ö ×�ÙUÚ Û ��� � Öã×SÙUÚÙBä ÂBγÏÞÄ åÞåÞåÖ ×�ÙUÚä�Ù ÂBÎ8Ï8Ä �...
���� Ñ with
Öã×SÙUÚÍ Ê ÂBγÏ8ÄLÛ À Öã×�ÙUÚÊ Í ÂBγÏÞÄ&Ñ (2.27)
whereas
Ö ×QäÚis neither symmetric nor anti-symmetric:Ö ×�äÚ Û ���
ÖØ×QäÚÙ�Ù ÂBÎ8Ï³Ä Öã×�äÚÙBä ÂBγÏÞÄ åÞå8åÖ ×QäÚä�Ù ÂBÎ8Ï³Ä Ö ×�äÚä�ä ÂBγÏÞÄ...
���� å (2.28)
Theory 26
Finally, the coupling with the laser is given by (2.9). In the symmetric matrix � the diagonal elements are the dipole moments of the molecule on the correspond-
ing potentials and the off-diagonal elements are the transition dipole moments con-
necting the electronic states:
� �� ������������ ����������������� ��� �������� ��� �������...
��� "! with #%$&������� $'#'�������(� (2.29)
2.2.3 An alternative description: The diabatic picture
Starting with the adiabatic time dependent Schrodinger equation (2.25) the deriv-
ation of the diabatic representation follows mainly the paper by Baer [52]. Yet, in
opposition to the latter article the equations are formulated one-dimensional and
time-dependent including interaction with an electric field in the electric dipole ap-
proximation. The diabatic basis can be obtained from the adiabatic one by a unitary
transformation, )* , of the electronic wave functions +%, #�- :+ . #�- )*0/ +%, #�- � (2.30)
The kets + . #1- form the diabatic basis, and equation (2.30) defines the unitary trans-
formation )* of the adiabatic into the diabatic basis set. Both, the total wavefunc-
tion expressed in the diabatic basis,2�35476980: ; +=< �1>?� - A@ #CB # � 6980: ; ! >?� + . # � 698D: ; � - (diabatic) ! (2.31)
and the total wavefunction in the adiabatic basis,2�354'6980: ; +%< ��>?� - @ #FE # � 6980: ; ! >?� +%, # � 6980: ; � - (adiabatic) ! (2.32)
have to be identical. This defines the relation between the diabatic nuclear wave-
function B and the adiabatic one, E :
E * B ! (2.33)
From equation (2.33) it follows thatGG ��� E GG ��� * B H * GG ��� B ! (2.34)
and G �G � �� E G �G � �� * B HJI GG ��� * GG ��� B H * G �G � �� B � (2.35)
Theory 27
Substitution of the last three equations, (2.33), (2.34) and (2.35), in the adiabatic
representation of the time-dependent Schrodinger equation (2.25) yieldsK7LNMMPORQ S TVU LXWY[Z]\_^ ` M WMba W\ Q S c YdMMea \ Q MMba \ S cdQ M WMba W\ S fgcJh Q SU Y L WY9Z]\Ri jlknmo` MMea \ Q S cdQ MMea \ S f U L WY[Z]\Ri j W m Q S Ugp q O?r0stvuZ Q STVU LRWY[Z]\ ^ Q M WMba W\ S cJh Q S U Y LRWY[Z]\ ` MMba \ Q c i jlknm Q f MMba \ S
U LXWY[Z]\w` Y i jxknm MMba \ Q c i j W m Q c M WMba W\ Q f S Uyp q O?r stvu Z Q S z (2.36)
By definition, the kinetic coupling terms are replaced by potential coupling in the
diabatic representation. It will be proved that, if Q fulfills the differential equation
(2.37): MMba \ Q c i jlknm Q T|{~} (2.37)
also the term Y i jlknm MMba \ Q c i j W m Q c M WMba W\ Q (2.38)
vanishes.
To prove the last statement, equation (2.37) is differentiated,M WMba W\ Q TVU MMba \ i jxknm Q U i jlknm MMba \ Q } (2.39)
and substituted into equation (2.38):Y i jlknm MMba \ Q c i j W m Q U MMba \�i jxknm Q U i jxknm MMba \ Q T|{ (2.40)
� i jxknm MMba \ Q c i j W m Q U MMba \ i jxknm Q T|{5z (2.41)
Inserting (2.37) into the last expression leads toU i jlknm i jlknm Q c i j W m Q U MMba \�i jxknm Q T|{~z (2.42)
It can be shown that the termU i jlknm i jlknm Q c i j W m Q U MMba \ i jxknm Q (2.43)
Theory 28
vanishes identically. The proof is given in section 2.9.3. It is therefore shown that,
if � satisfies the differential equation (2.37) the kinetic coupling terms � �x�n� and� ���'� , defined in equations (2.22) and (2.23), respectively, vanish and the time de-
pendent Schrodinger equation (2.36) is expressed as:�7����P� � � ��� � ��[����� � � ��e� �� � ��� � � �g� � �?�(�]� ��¡ � � � ¢ (2.44)
By multiplication of the last expression with � £ the final result is obtained:�'� ��P� � �¤� � ��[�]� � � ��b� �� � �d� £ � �¥ ¦�§ ¨© ªn«¬ � �g� � �?�®� � � � £ ¡ � �¥ ¦�§ ¨¯ ° ª±«=¬ � ¢ (2.45)
In the last expression � �A� £³²is the diabatic nuclear wavefunction,� £'� � �|� ´�µ � (2.46)
denotes the diabatic potential matrix which is in general not diagonal and� £ ¡ � � � ¡ � ´�µ � (2.47)
defines the diabatic dipole matrix.
2.2.4 Comparison between the adiabatic and the diabatic rep-
resentation
In the adiabatic representation the potential matrix is diagonal, whereas the kinetic
energy has coupling terms, � �l�n� and � �¶�'� . In the diabatic picture the kinetic coup-
lings are replaced by potential couplings, therefore, the diabatic potential matrix
has off-diagonal elements which couple the different states. Both representations
are equivalent in the physical sense [53] and are connected by the unitary matrix� that defines the transformation of the nuclear and electronic wavefunction, the
potential and kinetic energy and the transition dipole moment (see previous sec-
tion). In this work, the adiabatic potentials and transition dipole moments come
from quantum chemistry calculations and the kinetic coupling terms � �l�n� and � �¶�'�have been numerically evaluated from the ab initio data (chapter 3).
The physical interpretation of the adiabatic representation is straightforward:
The potentials and transition dipole moments are computed for fixed positions
of the nuclei using approximate methods for the solution of the electronic
Theory 29
Schrodinger equation (section 2.8). The kinetic couplings are only large around
avoided crossings which mark the break-down of the Born-Oppenheimer approx-
imation (= a strict separation of the electron and nuclear motion). The Franck-
Condon principle implies that electronic transitions are vertical transitions mean-
ing that the electronic (adiabatic) potentials do not change during the transition.
This principle allows the calculation of the absorption spectra.
In the diabatic case, the potentials and consequently the excitation energies dif-
fer from the adiabatic ones. Therefore, it is not clear how the diabatic excitation
energies can be interpreted (i.e. if they can be related to experiment). For that
reason the adiabatic representation in dynamics simulations was used throughout
this work. This was possible because the kinetic couplings are smooth enough
to use them as non-adiabatic couplings in the numerical calculations. Neverthe-
less, the use of the potential couplings (i.e. diabatic picture) is inevitable in all
cases where the kinetic couplings defined by (2.22) and (2.23) are very sharply
peaked around avoided crossings causing numerical difficulties. These computa-
tional problems of the adiabatic representation can be avoided by transforming into
the diabatic basis, where the kinetic couplings are replaced by potential coupling
functions which are usually smooth [54].
The problem of transforming the adiabatic into the diabatic representation has
been treated in recent studies [55, 56, 57]. The diabatic representation has been
used to treat atom-atom collisions [54, 58], atom-molecule interactions [52, 59],
photodissociation of OH [60], photodesorption [61] and photodissociation of or-
ganometallic compounds [34], amongst some applications. In the latter two art-
icles by Saalfrank and coworkers a diabatization procedure is applied which makes
use of the fact that the kinetic coupling terms can be approximated by Lorentzians
[62] if numerical results are not available.
2.2.5 Born-Oppenheimer dynamics
In the Born-Oppenheimer approximation [63] the kinetic coupling terms ·¹¸lºn» and· ¸�¼'» in equation (2.21) are neglected. The first term of the left-hand side of equa-
tion (2.21) defines the Born-Oppenheimer dynamics (without coupling due to the
electric field). Within this approximation the one-dimensional time-dependent
Schrodinger equation is given by:
½7¾N¿¿bÀ9ÁÃÂ7Ä�Å�Æ9Ç À?ÈÊÉ ËÌÌÌÍÏÎ ¾ ¼Ð[Ñ Æ ¿ ¼¿ Å ¼ÆNÒ�Ó Â�Ä�Å�Æ ÈÔ Õ�Ö ×Ø�Ù ¸�Ú�Û�»Ü�ÝÝÝÞ ÁÃÂ'Ä�Å�Æ[Ç À?騧 (2.48)
Theory 30
The infinitesimal change of the electronic wavefunction when changing the inter-
nuclear distance, àà(á�â�ã%äæå�ç , vanishes when the electrons follow the nuclei instantan-
eously. The Born-Oppenheimer approximation breaks down in the case of avoided
crossings of electronic potential curves of the same symmetry and spin. The kinetic
couplings are the larger the smaller the reduced mass is, i.e. when light particles are
involved (for example hydrogen atoms), and the higher the momentum, èêé7ëyàà®á âíì å ,of the particles is.
2.2.6 Population and probability of dissociation
The normalization condition for a wave function in one dimension, for example in
the Mn-CO stretching coordinate î�ï , isð±ñ ã ñ ç á âóòVôíõ (2.49)
In the adiabatic picture the total wave function, ã ñ÷ö�ø?ù ç , is expanded in terms of the
electronic wave functions ã=äûú1ç :ð î�ï ã ñ÷ö�ø?ù ç ò|ü ú ì ú ö î�ï[ý ø?ù ã=äûú1ç?þ±ÿ õ (2.50)
Under the assumption that the electronic wave functions are orthonormal,ð äûú�ã%äæå�ç ò�� ú=å ý (2.51)
the normalization condition gets:ô ò ü ú ������ � î�ï ì�ú ö î�ï�ý ø?ù ì ú ö î�ï�ý ø?ù õ (2.52)
The population ú of the é -th electronic state can be then defined as the part of the
total norm (which is equal to one) calculated by: ú ö�ø?ù ò � ���� � î�ï ì�ú ö î�ï[ý ø?ù ì ú ö î�ï[ý ø?ù õ (2.53)
Furthermore, the vibrational part of the nuclear wave functions ì�� ú� ú can be expan-
ded in terms of the vibrational eigenfunctions � � ú� ��� (a numerical method for calcu-
lating them is given in section 2.6):ì � ú� ú ö î�ï[ý ø?ù ò ü ����� ��� ö1ø?ù � � ú� ��� ö î�ï ù õ (2.54)
Then the population of the é -th electronic state is given by the sum of the popula-
tions on the different vibrational eigenstates: ú ö�ø?ù ò � ���� � î�ï ü�� ��� �� � ö�ø?ù � � ú� ��� � ö î�ï ù ü ����� ��� ö1ø?ù � � ú� ��� ö î�ï ù
Theory 31���������! ���"$#���&%('*) " ��,+ %�'*).-0/�11 2 ,3�4�5���6%�7$89) 2 ,3�4 ��:%�7;8;)*<=7;8?> (2.55)
Assuming that the eigenfunctions form an orthonormal set,-0/�11 2 ,3�4�5��� %�7;8$) 2 ,3�4 �� %�7$89)&<=7;8 ��@ ���A ��&B (2.56)
equation (2.55) becomesC 3 %�'*) � � ���"D# ��*%�'*) " �� %('*) � � ��FEG" �� %�'*) E H (2.57)
and the population of the I -th vibrational level of the J -th electronic state is defined
by the coefficients " 3LK : C 3M %('*) �NEG" 3M %�'*) E H > (2.58)
Similarly, in one dimension it is possible to define the probability of dissociation
of the system in a certain electronic state as the part of the population which lies
between a defined dissociation point O and the asymptotic region of the considered
potential: CQP 4�R3N%�'*) � -S14 T #3 %�7$8 B '*) T 3 %�7;8 B '*)*<=7;8U> (2.59)
Usually a pragmatic, yet reasonable, choice for the dissociation point O is a value
of three times the equilibrium distance of the reaction coordinate. An alternative
way to define O is to chose the point at which the orbitals of the fragments are pure,
meaning that no orbitals with contributions centered at both fragments at the same
time are present. If the wavepacket or a part of it can move towards dissociation
(and is not trapped by a barrier) the probability of dissociation yields a rising signal
from which a decay time V of the excited state can be calculated. It must be pointed
out that a computation of V using (2.59) only makes sense if the applied laser pulses
are much shorter than V . In this work, dissociation probabilities and corresponding
life times have only been calculated for infinitesimal short @ -pulses (see section
4.3).
Theory 32
2.3 Rotational averaging
In this work freely rotating (i.e. not oriented) molecules are considered. The di-
pole moment and the electric field are defined in two different coordinate systems:
The space fixed coordinate system is chosen such that the W -axis is parallel to the
electric field vector X Y(Z*[ , whereas the dipole moment is given in the molecule or
body-fixed coordinate system with the z-axis pointing in a different direction (e.g.
parallel to the dissociation coordinate [41]). The rotation matrix (2.61) connects
the body-fixed coordinate system (which rotates with the molecule) with the space
fixed one. In the semiclassical dipole approximation the interaction of two elec-
tronic states \ and ] is given by^`_ba Y�c$d efZ*[,g a Y�c;d efZ*[ihkjlX Y�Z*[nmpoq msr _ba Y�c$d [tm9g a Y�c$d efZ*[ne (2.60)
where oq is the rotation matrix [64]oq h uvw qyxzx qyxz{ qyxn|qy{*x qy{&{ qy{*|q}|�x qQ|�{ qQ|,| ~;�� (2.61)
connecting both coordinate systems, in this work, approximated by rotational av-
eraging of oq . To derive the working equations three electronic states will be con-
sidered ( \th���e9��en� ). Only transitions from the ground state ( \�h�� ) to the excited
states ( \6h ��en� ) are taken into account. Therefore, the transition dipole matrix
reads as: � r h uvw � r �f��r ���r ��� � �r ��� � �~;���� (2.62)
The Born-Oppenheimer dynamics of this three-state-system is governed by the fol-
lowing set of three coupled differential equations:\,�Q�� Z g _ Y�c;d efZ*[ih���j � ���r�d � �� c �d���� _ Y�c$d [&�`g _ Y�c$d efZ*[�j�X Y�Z*[Dm oq�� a r _ba Y�c$d [�m�g _ Y�c;d efZ*[ne\�h ��e9��en� � (2.63)
Consider an ¡ -polarized laser pulse which propagates along the z-axis. Then,^`_ba Y�c;d efZ*[,g a Y�c$d efZ*[ihj uw X x e¢X {£U¤;¥U¦§ � e¨X |£U¤;¥U¦§ � ~� m©uvw qyxzx qyxz{ qyxn|qy{&x qQ{&{ qQ{*|qQ|�x q}|�{ qQ|,| ~;�� m©uvw r _ba,ª xr _ba&ª {r _ba&ª |
~;�� m�g a
Theory 33
«�¬�¯®&°²±�³y°z°=´!®*°µ±?³Q°z¶U´z®&°l±?³y°n·n¸t±º¹»¼ ½�¾b¿&À °½Á¾b¿&À ¶½�¾G¿&À ·Â;ÃÄ ±;Å ¿
«k¬Æ¯®*°l±?³y°z°©± ½�¾b¿&À °©±�Å ¿tÇ ®*°l±?³Q°z¶:± ½�¾b¿&À ¶È±�Å ¿�Ç ®&°l±?³y°n·É± ½�¾b¿,À ·É±�Å ¿ ¸�Ê (2.64)
Let us now assume that the system is photodissociated by a Ë -pulse, i.e. we use® �Ì*¸i« ® for Í}Î Ì Î0Ë Ì® (Ì*¸i« Í else, (2.65)
where Ë Ì is a sufficiently short time interval, such that it allows one to approximate
the time derivative in equation (2.63) by a finite difference:ÏÏ Ì Å ¾�Ð Å ¾ Ë Ì*¸t¬FÅ ¾ (Ìi« Í ¸Ë Ì Ê (2.66)
Using the last expression with the initial conditions,ÅÒÑU�Ìi« Í ¸i«�ÓÔÑ�Õ9´Å�ÖD�Ìi« Í ¸�« Í ´Å×��Ìi« Í ¸�« Í ´where ÓÔÑ�Õ is the vibrational and electronic ground state, in equation (2.63) one
obtains (neglecting the term Ø ½ Ñ�Ñ ):ÙÚ&Û�Ü ¬ Û ×Ý ½pÞ Ï ×Ï.ß ×Þ ÇSà ÑzáâÓÔÑ�Õ Ð ÅÒÑU Ë Ì*¸�¬�ÓÔÑ�ÕË Ì ã ÅÒÑU Ë Ì*¸i«�ÓÔÑ�Õi¬ Ë Ì�± ÚÛ ±?ä©Ñ�ÕɱUÓÔÑ�Õ Ð ÓÔÑ�Õ9´ÙÚ,Ûæå ¬²® ±pç³ ± ½ Ö�Ñnè ÓéÑ�Õ Ð ÅtÖ; Ë Ì*¸Ë Ì ã ÅtÖ; Ë Ì*¸i« Ë Ìt± ÚÛ ±?® ±�ç³ ± ½ Ö�Ñ ±UÓéÑ�Õ;´ÙÚ,Û å ¬²® ± ç³ ± ½ ×�Ñ è ÓéÑ Õ Ð ÅÒ×? Ë Ì*¸Ë Ì ã ÅÒ×U Ë Ì*¸�« Ë Ì�± ÚÛ ±?® ± ç³ ± ½ ×�Ñ ±UÓéÑ Õ Ê (2.67)
Substituting equation (2.64) into equation (2.67), the population at time Ì�« Ë Ì is:
ForÚ « Íëê�ì Ñ? Ë Ì*¸ Ð�íïî ß Þ ÓéðÑ�Õ ÓÔÑ�ÕÈ« Ù Ê
ForÚ « Ù ´ Ý ê�ì ¾ Ë Ì*¸ Ð
Theory 34ñfòó²ôUõ÷ö=ø;ù�ú�û ózó û ózóýü þ�ÿ���� ó������ ü ò û ózó û ó�� ú þÁÿ���� ó ��������� ú þ�ÿ���� ��������� û ózó û ó�� ú þ�ÿ���� ó���������� ú þ�ÿ���� ��������� û ó�� û ó���ü þ�ÿ���� ����� � ü ò û ó�� û ózó ú þ�ÿ���� ����� � � � ú þ�ÿ���� ó ��� � � û ó�� û ó�� ú þ�ÿ���� ����� � � � ú þ�ÿ���� ����� � � û ó�� û ó�� ü þ�ÿ���� ������� ü ò û ó�� û ózó ú þ�ÿ���� ��������� � ú þ�ÿ���� ó�������� û ó�� û ó�� ú þ�ÿ���� ��������� � ú þ�ÿ���� �������������Averaging over angles:û ózó û ó���� û ózó û ó���� û ó�� û ó����! #"û ózó û ózó�� û ó�� û ó��$� û ó�� û ó����&%' " (2.68)
leads to(*) ú,+ - �/. ñ òó ô õ ö=ø;ù %'10 ü þ ) ��� ó���� � ü ò ü þ ) ��� ����� � ü ò ü þ ) ��� ����� � ü ò�2 (2.69)
( ò ú3+�- �*. ñ òó ô õ ö ø$ù %'40 ü þ ò ��� ó ��� � ü ò ü þ ò ��� ����� � ü ò ü þ ò ��� ����� � ü ò�2 � (2.70)
Therefore, the initial excited state wave packet is given by the ground state wave
function ��� � multiplied by the corresponding transition dipole moment func-
tion. After the + -pulse, the populations of the electronic states do not change
and the wave packet in each excited state is governed by equation (2.63) (Born-
Oppenheimer dynamics).
2.4 Time dependent calculation of absorption spec-
tra
According to Heller [65] the total absorption spectrum 57698:6 ú<; � , which measures
the capability of the molecule to absorb radiation with a frequency ; is obtained
from the Fourier transformation of the total autocorrelation function,57698:6 ú<; �/= õ!>@?? A ÿCBED � ��F7G�HJI 6LK G�M 698:6 úN- � öO- " (2.71)
where P � � is the energy of the vibrational ground state of the electronic ground
state and Q ; is the energy of the absorbed photon [41]. The total autocorrelation
functionM 698:6 ú<- � is defined as the sum of the individual autocorrelation functionsMSR ú<- � of the electronic states T .M 698:6 ú<- �/�VU R MSR úN- �*�WU RYX,Z R ú<- �W [�?ü Z R ú<- ��\ (2.72)
Theory 35
The autocorrelation function describes the overlap between the wave function]^`_<a�b and the initial wavefunctions ]c^d_<a*eWfOb as a function of time t. In this work,
the initial wave function of the excited states g is calculated by multiplying the vi-
brational ground state wave function of the electronic ground state hjilk�mn�o with the
transition dipole moment that connects the ground state with the state p :q ] ilk�m^ _<a n eWf[b�r*ets n ^vu w�x:yvx{z}| q h ilk�mn o r (2.73)
A derivation of equation (2.71) is given in the appendix A.
2.5 Pump-probe ionization spectroscopy
Two steps are involved in a pump-probe ionization scheme. In a first step a laser
pulse, called pump pulse, excites the molecule. In a second step, after a certain
time delay, another laser pulse, the probe pulse, is employed to ionize the molecule.
The theoretical pump-probe signal of the parent ion is determined here as the part
of the population which is trapped in a certain ionic state, whereas the yield of the
parent and the daughter ions correspond to the non-dissociative and the dissociat-
ing parts of the ionic states population, respectively. The pump-probe ionization
spectrum reflects the nuclear dynamics of the electronic excited state as a function
of the delay between the pulses.
In this work the vibrational ground state of the electronic ground state, hjilk�mn�o ,
is chosen as the initial wavefunction. Since this is an eigenstate of the system,
the corresponding expectation values (for example of the position operator) do
not change with time (stationary state). The pump pulse, however, creates a wave
packet in the excited state which is a coherent superposition of stationary states
[41]. Being not an eigenstate, the wave packet moves governed by the time-
dependent Schrodinger equation. Consequently, the wave packet is located at dif-
ferent areas of space at different times. That is the reason why the ionization prob-
ability will depend on the delay time [66].
The electron removed during the ionization process has a continuous spectrum
of allowed kinetic energies of the detached electron [67, 68, 69, 70, 71, 72, 73, 74].
However, for a numerical treatment the kinetic energy spectrum has to be dis-
cretized. As pointed out in refs. [75, 76] the ionization occurs very rapidly with
approximate conservation of the nuclear kinetic energy and the contribution of
a single optimal selected photon energy already defines the dominant features
of the total transient ionic signal. Hence, in the approach of this work it is as-
sumed that the electron removed during the ionization process has zero-kinetic-
energy (ZEKE). The laser-induced transitions to ionic states are then treated in the
Theory 36
same way as to neutral states. However, it is important to avoid unphysical back-
transformation (dump) of ionic state population by the applied lasers. This can be
achieved by means of small intensities of the pump and (more importantly) the
probe pulse. (e.g.: In this work the intensities were chosen such that the pump
pulse transfers about 10 ~ of the ground state population to an electronic excited
state and the probe pulse produces an ionic state population of less than 1 ~ .)
2.6 Calculation of the initial wave function: The
Fourier Grid Hamiltonian (FGH) method
Within the Born-Oppenheimer approximation (equation (2.48)) the one-
dimensional time-independent nuclear Schrodinger equation is:�S���d����S�4� ���� ���������� � ������ ��� ����������N��� j¡ ��¢¡ � � ����*£t¤
¡ � �¡ � � �
����¥(2.74)
The nuclear wavefunctions ¡ ��¢¡ � � ����
are the vibrational eigenfunctions of the
Hamiltonian ¦ �l� � ��� . The eigenvalues¤¡ � are the allowed total energies (neglect-
ing the rotational and translational contributions) of the one-dimensional system.
Both, the energies¤¡ � and the wavefunctions ¡ ��¢¡ � of the § -th electronic state are
labeled by the index ¨ � , the vibrational level corresponding to state § .By solving the nuclear Schrodinger equation (2.74) for the electronic ground
state potential ��© � � ��� the stationary states ¡�ª � ����
and corresponding energies¤¡�ªin the Born-Oppenheimer approximation are obtained. The Fourier Grid Hamilto-
nian method [77], whose principles are described in this section, has been em-
ployed to solve the above equation. An alternative method to get at least the en-
ergetically lowest eigenfunction, not applied in this thesis though, is the so-called
”propagation in imaginary time” from Kosloff and Tal-Ezer [78].
In equation (2.74) ¦ � is represented in coordinate space:« ��¬�d@®¦ � � ��¯°£ « � ¬�d ®± � � ��¯ � « � ¬�`�®��� � ��¯�¥ (2.75)
The potential operator �7� is a function of the position operator®²
with eigenvalues� � and eigenvectors � ��¯ : ®² � � ��¯}£ � � � ��¯�¥ (2.76)
Its ”matrix elements” in the coordinate representation are then [51]:« � ¬� �®�7� � ��¯°£ ����� � ����³ � � ¬� � � ����¥ (2.77)
Theory 37
The action of ´µ�¶ in coordinate space is a simple multiplication of the wave function·¹¸Nº�»�¼with the potential energy function
µ�½ ¸,º�»�¼, as expressed in equation (2.74).
On the other hand, the matrix elements of the kinetic energy operator ´¾¿¶ in the mo-
mentum representation are given by,À{Á`Â{à ´¾�¶ ÃÄÁ7Å*ÆÈÇ Á@ÉÊÌË »`Í ¸ Á`Â`ÎÏÁ ¼�Ð(2.78)
whereÃÄÁ#Å
are the eigenvectors andÁ
the eigenvalues of the momentum operator´Ñ : ´Ñ ÃÄÁ#Å}ÆtÁcÃÄÁ#Å�Ò(2.79)
The completeness relation holds for the coordinate and momentum eigenstates:ÓYÔÕ Ô Ã º�» Å�À º�» à Ö[×ØÆ ´Ù (2.80)
and Ó ÔÕ Ô ÃÄÁ7Å�À{ÁcÃÚÖ`ÁÛÆ ´Ù Ò (2.81)
Inserting (2.81) and (2.77) in equation (2.75) yieldsÀ º » à ´Ü�¶ à º�» Å*Æ ÓYÔÕ Ô À º » à ´¾�¶ ÃÄÁ7Å�À{Á�Ý º�» Å�Ö@Á�Þ µ�¶ ¸,º�»�¼ Í ¸Nº » Î º�»�¼Æ Ó ÔÕ Ô À º » ÃÚÁ7Å Ç Á ÉÊÌË » À:Ácà º�» Å�Ö@ÁßÞ µ�¶ ¸,º�»�¼ Í ¸Nº » Î º�»�¼
(2.82)
The transformation matrix elements between the coordinate and the mo-
mentum representation are À º » ÃÄÁ7Å}Æ Ùà Ê�á$â ¶�ã�ä:åæ (2.83)
and À{Ácà º�» Å°Æ Ùà Ê�á$â Õ ¶�ãvä æ Ò (2.84)
Therefore, (2.82) getsÀ º » à ´Ü�¶ à º�» Å*Æ ÙÊ�á ÓYÔÕ Ô â ¶�ã�çEä åæ Õ ä ævè Ç Á@ÉÊÌË » Ö`ÁßÞ µ7¶ ¸Nº�»�¼ Í ¸,º » Î º�»�¼ Ò(2.85)
The last equation is the heart of the FGH method. To get the eigenfunctions of´ÜÛ¶ the position operator ´é is discretized by substituting the continuous coordinate
valuesº�»
by a discrete setº�ê»
:º ê» Æ!ëSì º�»�Ð ëíÆ Ù Ð ÒîÒLÒ Ð�ï Ò(2.86)
Theory 38ðòñ�ódenotes the spacing in coordinate space and N is the number of grid points.
Finally, the following expression for the matrix elements of ôõ�ö is obtained [77]:õÛ÷ùøûú üðÛñ�óÛý þÿ÷�� � þ� ö�÷������ � ø��� þ� ��� ÷����7ö�� ñ ó���� �ø�� � (2.87)
with � ÷�ú ! �"$# � �&% ð(' � � � ð(' ú "$)+* � ðÛñ�ó$,(2.88)
Diagonalizing the N � N matrix of the Hamilton operator (2.87) yields the eigen-
vectors and eigenvalues of ôõÛö on the chosen grid.
Theory 39
2.7 Propagation schemes for the time dependent
Schrodinger equation
The commonly used propagation schemes are critically analyzed in a review article
by Leforestier et al. [79] and recapitulated by Balakrishnan et al. [80]. Nonethe-
less, for the sake of completeness, the basic equations are presented in this section.
In addition, it will be shown how the kinetic coupling terms introduced in section
2.2 can be included in the second order differencing (SOD) algorithm.
The Hamilton Operator in equation (2.2) is a sum of the kinetic energy oper-
ator -. and the potential energy operator -/ . The evaluation of -/10 in coordinate
space is straightforward, since -/ is diagonal in this representation and its action
consists of a simple multiplication. The bottleneck of all quantum time-dependent
propagations is the calculation of -.20 , which is diagonal only in the momentum
representation. Kosloff and Kosloff [81] introduced the fast Fourier transforma-
tion technique (FFT), in combination with a second order differencing scheme, to
evaluate the action of the kinetic energy part of the Hamiltonian in the momentum
space. Using the fast Fourier transformation algorithm the second derivative of the
wave function is calculated in three steps:
1. Inverse Fourier transformation (IFT) to momentum space:30547698;:=< >?$@BADCE C 054GF�HI8�J E�KMLON&PIQ F�H (IFT) R2. Multiplication with S 6�T (
6= wave number),
3. Fourier transformation (FT) to coordinate space:054&FIH�8;:=< >?$@ ADCE C 3054U698�J KMLON&P Q 6 (FT) VIf the Hamilton Operator is time-independent, the formal solution of (2.1) is [51]:W 054GX�8�YZ: -[\4]X R X�^_8 W 054GX�^`8�YZ:aJ EcbdZef;g�h E hjilk W 0(4]X�^`8OY R (2.89)
where[m4GX R X�^�8 is the time-evolution operator. An electric field makes the Hamilton
Operator time dependent. If, however, the time step is chosen so small that the
change of the electric field is negligible, the Hamilton operator can be treated time
independent and (2.89) is still valid. In this work the split operator [82, 83, 84] and
the second order differencing scheme [81] were used to solve the time dependent
Schrodinger equation numerically. Since both methods require short time steps
(which are also needed for handling time-dependent Hamiltonians) they are a good
Theory 40
choice in the presence of molecule-laser-interactions. Both algorithms are unitary
and norm preserving. The SOD method conserves the energy whereas the split
operator does not. Moreover, the split operator technique cannot be applied when
space-momentum mixed terms - for example kinetic coupling terms - appear in
the Hamiltonian [79]. Therefore, in this work, the split operator was used when
the kinetic couplings were neglected as a first approximation, while when these
couplings were included, the time dependent Schrodinger equation was integrated
by means of the second order differencing scheme.
2.7.1 Second Order Differencing (SOD)
One way of solving the time dependent Schrodinger equation numerically is to ex-
pand the time dependent wavefunction in a Taylor series [85, 86]:n5oGprqtsmp�u;vwn5oGp�urqxs\p�y2zz p n5o]p�urq smp�{|�} y~z {z p { n5oGp�urq smp����} y2z �z p � n5oGp�u�y�y�y(2.90)
or n5oGp���smp�uZvwn(o]p�u���smp�y2zz p n5oGp�urq smp {|�} y�z {z p { n5o]p�u�� smp ���} y~z �z p � n5oGp�u�y�y�y��(2.91)
Subtracting equation (2.91) from equation (2.90) yieldsn5oGp�qtsmp�u��xn5oGp���smp�u�v | y�smp�y�zz p n5o]p�urq | y smp����} y2z �z p � n5oGp�u�y�y�y�� (2.92)
The resulting propagation scheme with a third order error insmp
isn5oGp�qtsmp�u;vwn5oGp���smp�u�q | y�smp�y zz p n5oGp�urqt����smp �`� � (2.93)
This algorithm requires both the initial wave functionn(o&��u
and the wave function
at the first stepn5oUsmp�u
. The initialization scheme used in this work is the Runge-
Kutta propagation with the same accuracy (second order) as the SOD method:n(o&smp�uZvwn(o&��u�qxs\p�y2zz p n5o&��u�q smp {|�} y2z {z p { n5o&��u_� (2.94)
According to (2.1) the term �_���� reads, in atomic units,z nz p v������� nwv¡ ���D¢|£�¤ z {z¦¥ {¤ �§��¨©o ¥ ¤ u�� ¢£�¤�ª¬«®U¯ o ¥ ¤ u�y ¡ ���l° zz¦¥ ¤9± qx�t¢|£�¤�ª¬« { ¯ o ¥ ¤ u ± n\² (2.95)
Theory 41
where the potential operator ³´ contains the coupling with the electric field. The
kinetic energy operator ³µ and the momentum operator ³¶ act in momentum-space
as multiplications with ·¸�¹_º�» ¸ and » respectively. The following scheme can be
applied:
1. Inverse Fourier transformation to momentum space:¼5½&¾�¿IÀ IFTÁàļ5½ » À_Å2. Calculate the action of ³µ and ³¶ in momentum space:Æ�Ç�È�ÉÊÌË ºÎÍ ÊÁàÁ~ÏÑÐ ³µ ļ5½ » ÀlÒ\ÅÈ ÍÁ� Р³¶ ļ5½ » À Ò Å3. Fourier transformation of ³µ2¼ and ³¶�¼ to coordinate space
FTÁàÁ�ÏÑÐ ³µ~¼5½&¾I¿ÓÀlÒ (term I)Å
Ð ³¶Ô¼5½&¾�¿�À Ò Å4. Calculate the action of all other operators in coordinate space:¼5½G¾�¿�À Æ�ÇjÕrÖØ× º�ÙÁ� Á�Ï ´©½G¾�¿�ÀO¼5½&¾I¿�À (term II)
ÅÐ ³¶�¼5½&¾I¿�À Ò Æ ÉË ºÚÜÛ É�Ý ÖÞ× º ÙÁ� Áàßá ¿ µ Ö · Ù ½&¾I¿ÓÀ Ð ³¶�¼5½&¾I¿�À Ò (term III)
ż5½&¾�¿�À Ç ÉÊÌË ºâÚ Û Ê Ý ÖØ× º�ÙÁÃÂ Ï ßã á ¿ µ Ö ¸�Ù ½&¾�¿�ÀO¼5½G¾�¿�À (term IV)
Å5. Sum up terms I to IV to get Á�Ï�ä ¼wåàæ_çæ�è .In accord with the uncertainty principle, the time step must not exceed the crit-
ical time step émê�ë]ì Ç è å íî�ï�¿�ð Å (2.96)
where the maximum energy î�ï�¿lð¬åñµòï�¿�ðôót´¦ï�¿lðÎõ(2.97)
In the last equation´¦ï�¿lð
is the maximum potential energy andµÃï�¿lð
the maximum
kinetic energy of the grid defined asµÃï�¿�ðöå í ¸O÷r¸ã á ¿½ émø À ¸ õ (2.98)
For practical calculations a time step five times smaller than the critical one is re-
commended [79].
Theory 42
2.7.2 Split Operator
In this method the time propagation operator ùú\ûGü`ýOü�þ_ÿ of equation (2.89) is approx-
imated as [82, 83, 84]:ùúmûGü����mü`ýOü�ÿ�������� ����� ������ �� ����� ����� ����� ���� �� ����� �������mü� "!�#(2.99)
The propagator (2.99) leads to an error of third order in the time step as a con-
sequence of the noncommmutability of the kinetic and potential energy operators.
A slightly larger time step than that of the SOD method specified in section 2.7.1
can be used [80].
This method involves the following steps: The wave function is transformed
to momentum space and multiplied with�%$'&+û�(*),+-/.10�2�3 �mü�ÿ . After transforma-
tion to coordinate space it is multiplied with�%$'&+û4(5)+76 û98%:�ÿ;�mü�ÿ . The resulting
wave function is again transformed to the momentum space and multiplied by�%$'&+û4( ),+-/. 0 2<3 �mü�ÿ . A following Fourier transformation to coordinate space com-
pletes one evolution step. Since the factor�%$'&�û4( ),+-/. 0 2�3 �\ü�ÿ is independent of the
step of the propagation, the left and the right part in (2.99) of two successive
propagation steps can be combined.
2.8 Solution of the electronic Schrodinger equation
using ab initio and DFT methods
Deriving the adiabatic representation in section 2.2 it was assumed that the solu-
tions of the electronic Schrodinger equationù=?>A@CBEDGFIH��KJ >L@F BMDGFNHIý(2.100)
whereù=?>L@O�P( QR :TSVUXW 3Y[Z > �?:�� \]_^a` þcb ( QR ) Sd:*eRf SVU g f � 3B h : (ji f B � QR :lk<m � 3B h : (nh m Bpo (2.101)
is the electronic Hamiltonian in Cartesian coordinates, are known. In practice only
a limited number of electronic eigenfunctions are calculated and used as a basis to
describe the adiabatic behaviour of the molecule. How many electronic wave func-
tions are needed depends on the process to be described. In this work the photo-
dissociation under a femtosecond laser pulse excitation is investigated; therefore,
a sufficient representation must contain all states which are, directly or indirectly,
significantly populated after the applied pulse.
Nowadays, many sophisticated methods for an approximate solution of (2.100)
exist. In this section those which have been applied in this work will be explained
Theory 43
in subsections 2.8.1 (ab initio methods) and 2.8.2 (DFT methods). Before, some
general remarks about the applied strategy concerning the quantum chemistry cal-
culations will be made.
The first step of modern quantum chemical applications is usually to determine
the geometry by an optimization of all degrees of freedom. Compared with single
point energy calculations the geometry can normally be obtained at a rather low
level of treatment [87], (e.g. in this work the CASSCF method was used for geo-
metry optimizations but the MS-CASPT2 - a more sophisticated method - was em-
ployed for energy computations.) In a second step, the vertical excitation energies
are calculated at the optimized geometry. (Here, MS-CASPT2 and TD-DFT cal-
culations have been performed.) These energies can be compared with the exper-
iment and the accuracy of the applied quantum chemistry method can be judged.
For a correct description of the quantum dynamics the accurate computation of po-
tential energy curves, which are obtained in the adiabatic representation (e.g. equa-
tion (2.21)), is an essential requirement. The potential energy curves of big mo-
lecules like organometallic compounds can only be described by highly-developed
quantum chemistry methods in a restricted number of degrees of freedom (because
of computational cost). Usually, only the reactive coordinates leading to dissoci-
ation are taken into account (here, the Mn-CO q4r stretching coordinate). Besides,
on an ultrafast time scale the rest of the molecule is supposed to stay at a fixed
geometry, supporting that very few degrees of freedom are necessary and IVR can
be neglected. The choice of the reaction coordinates is guided by the structure of
the primary products which can be obtained via geometry optimizations of the mo-
lecule in the ground state and relevant electronic excited states.
2.8.1 Standard quantum chemical (ab initio) methods
Standard quantum chemical (ab initio) methods are based on a Hartree-Fock (HF)
[88] treatment. The molecular ground state sEtGu1v of stable molecules (if the system
has an even number of electrons) is usually well-described by a single closed shell
Slater-determinant, s wxulv , containing the y (= total number of electrons) energetic-
ally lowest spin orbitals z (determined in a HF calculation):
sMtGu"v|{�}�~ q4�I�Lq�����V� s w?u1v�� �� y�� ���������z���� �N� z��[� ��� �I�I� z��d� ���z����L� � z��[�A� � �I�I� z��d�A� ��%�I� �%�I� �I�I� �I�I�z����9y � z��N��y ���I�I� z�����y �
���������I� (2.102)
where the rows are labeled by electrons and the columns are labeled by spin or-
bitals. Using this ansatz for the electronic wavefunction to minimize the elec-
tronic energy ���A�� variationally leads to the Hartree-Fock equations. In a Restricted
Theory 44
Hartree Fock (RHF) calculation [88] the spin orbitals � are a product of the spatial
orbitals � (depending on the spatial coordinate ) and the spin function (depending
on the spin coordinate ¡ ), ¢ or £ :�¥¤� ¦;¡G§©¨«ª ��¤� §�¬N¢¥¤¡G§ or��¤� §�¬�£�¤¡G§ .(2.103)
With this ansatz the spin functions ¢ and £ can be integrated out. For a numerical
treatment the remaining spatial orbitals � are linearly expanded in a finite basis set®7¯"°N±of dimension M: �O²�¤� §³¨ ´µ °·¶ ² °�¯"° ¤� §1¦ (2.104)
leading to the Roothaan-Hall equations [89]:¸ ¹ ¨�º ¹ » ¦ (2.105)
where¸
is the Fock matrix, S is the overlap matrix,¹
is coefficient matrix and»
is the orbital energy matrix. The iterative solution of equation (2.105) is called the
Self-Consistent-Field (SCF) procedure.
The difference between the exact solution of the electronic Schrodinger equa-
tion (2.11), ¼ (Schr.), and the Hartree-Fock-limit energy, ¼ (HF-Limit), which is the solu-
tion of the Hartree-Fock equations [88] when using a complete basis expansion, is
defined as the correlation energy ¼¾½9¿AÀAÀ :¼¾½9¿AÀAÀ�¨�¼ (Schr.) Á ¼ (HF-Limit)  (2.106)
The obvious way to account for the correlation energy is to use Configuration Inter-
action (CI). For a given basis set a Full Configuration Interaction (FCI) calculation
constitutes a benchmark by which computations of the correlation energy with the
same basis set can be judged, i. e. ”full CI is the best that one can do” [88]. If the
basis set reaches completeness, the FCI result will be the exact solution ¼ (Schr.). The
FCI expansion of an electronic state reads as [88]:ÃEÄGÅ�ÆdÇ1È ¨ÊÉ"Ë ÃMÌ Ë ÈaÍ µCÎ À É À Î ÃMÌ ÀÎ È�Í µÎ"Ï<ÐAÑ À Ï�Ò É À ÒÎ4Ð ÃMÌ À ÒÎ4Ð È�Í ÂÓÂÔÂO (2.107)
In equation 2.107,ÃMÌ Ë È stands for the ground state configuration and
ÃMÌ ÀÎ È denotes
a single excitation, i.e. a Slater determinant where the spin orbital � Î which is oc-
cupied in the ground state is replaced by the unoccupied (=virtual) spin orbital �aÀ .Similarly,
Ã Ì À ÒÎ;Ð È is a doubly excited determinant where the orbitals � Î and � Ð are
Theory 45
replaced by the virtual orbitals Õ�Ö and Õ�× . The number of Ø -tuply excited determ-
inants for n electrons and 2 M spin orbitals, M being the number of spatial basis
functions Ù , is given by ÚÜÛØÞÝ
ÚKßàÛ ájâØ Ýäã (2.108)
From this formula it is clear that the number of configurations in the expansion
(2.107) grows very rapidly with the number of electrons and basis functions mak-
ing the FCI method only applicable for very small molecules and reduced basis
sets. Only truncated CI methods like CIS or CISD, where Single (S) and Double
(D) excitations are considered, can be used in general, but they suffer from size-
consistency (e.g. a CISD calculation of two H å molecules separated by a large dis-
tance (say 100 A) does not give twice the CISD energy of one H å molecule (which
is lower)). Furthermore, CIS gives poor excitation energies.
Due to their multiconfigurational character, electronic excited states can not
be described by a single Slater determinant and, therefore, a multiconfigurational
procedure is needed. A solution for computing electronic excited states or for
cases where a single determinant is not even a good zeroth order reference wave-
function is the so-called Multi-Configuration Self-Consistent Field (MCSCF) ap-
proach, which consists of a truncated CI expansion where not only the CI coeffi-
cients æGç in front of the Slater-determinants èMéêç9ëèEì¥íïîOð[îdñ�ë�òÊó ç æGç;èMéêç9ë (2.109)
are variationally optimized, but also the Molecular Orbital (MO) coefficients ôàçMõ in
the basis set expansion [90]. The practical problem lies on the choice of the relev-
ant configurations, èMéêçë . A popular solution consists of partitioning the molecular
orbitals in active and inactive spaces. This is the way how the selection of the con-
figurations is chosen in the Complete Active Space Self Consistent Field Method
(CASSCF) [91]. The inactive orbitals stay either doubly occupied or empty during
the calculation. Typically the active space orbitals consist of the highest occupied
and lowest unoccupied orbitals of a RHF wave function. In addition, calculations
of excited states of transition metal compounds containing metal atoms of the first
transition row have to deal with the problem of the öà÷ double shell effect: Two sets
of ÷ orbitals ( öà÷ and ø'÷ ) must be included in the reference space in order to obtain
accurate results [92]. Within the active orbitals a Full Configuration Interaction
(FCI) calculation is performed (figure 2.1). Therefore, the CASSCF method is on
the one hand a special MCSCF method, meaning that not only the CI coefficients
but also the orbital coefficients in (2.109) are optimized, and on the other hand a
Theory 46
active space
Allexcita-tions
Figure 2.1: All the possible excitations within the active space define the config-
urations in a CASSCF calculation.
special case of the FCI method, meaning that all possible excitations within the
active space are taken into account.
The correlation energy (2.106) can be divided into two different contributions:
The static and the dynamical parts. The static part of the correlation energy ac-
counts for the effect of allowing the orbitals to be partly singly occupied, like in the
CASSCF description, instead of forcing double occupation, like in the HF approx-
imation. This allows the description of near-degeneracy effects of molecular orbit-
als which are especially important for organometallic compounds. The remaining
correlation energy is the dynamic correlation which describes the correlated mo-
tion of the electrons. The latter part is normally taken into account by subsequent
perturbation treatment, CASPT2 or by the Multi-Reference Configuration Interac-
tion (MRCI) method [93]. Conventional CI methods like CISD consider only con-
figurations generated by exciting electrons from a single determinant [94], usually
the ground state RHF wave function. A MRCI calculation is based on a previ-
ous MCSCF treatment, for example CASSCF. The critical step in this method is
the choice of the reference wavefunctions, which has to be consistent along the
process investigated [95]. The externally contracted version of the MRCI method
called MR-CCI, introduced by Siegbahn [96], was applied in this work to calculate
the potential energy curves, transition dipole moments and kinetic couplings.
The CASPT2 method applies second order perturbation theory to a CASSCF
reference wavefunction [97, 98]. This treatment includes a large amount of the
Theory 47
dynamical correlation leading to very accurate results for excitation energies (nor-
mally the error is in the range 0.0-0.3 eV [92, 99]). However, the original version
of the CASPT2 method allowed the calculation of a single state at a time, which
made the description of curve crossing problems impossible. Recently, it was re-
placed by the Multi-State CASPT2 (MS-CASPT2) [100] method which make the
simultaneous treatment of more than one state possible.
Near-degeneracies in the zeroth-order Hamiltonian lead to the problem of in-
truder states. The solution of this problem is to increase the active space. However,
this is not always possible, since a larger active space increases the number of con-
figurations and therefore, the computational cost. A remedy to avoid intruders is
to introduce a level shift [101, 92], but then the amount of the correlation energy
included decreases.
For the chosen basis set and active space the CASPT2 method was not able to
describe dissociation correctly in CpMn(CO) ù . For that reason the MS-CASPT2
method was used in this work to calculate the vertical excitation spectrum, whereas
the CASSCF/MR-CCI method was applied for the calculation of the potential en-
ergy curves.
2.8.2 DFT methods
An alternative approach to the Hartree-Fock based methods is the Density Func-
tional Theory (DFT) which is based on the Hohenberg-Kohn theorems [102]. The
first Hohenberg-Kohn theorem (HK-I) states that there is a one-to-one mapping
between the external potential ú (i.e the Coulomb attraction of an electron by all
nuclei), the electron density û and the ground state wave function üGý :û?þÿú?þ üGý (HK-I) � (2.110)
This implies that all properties are functionals of the ground state density due to the
fact that they are calculated as expectation values of operators for the state vectorüGý � û�� corresponding to the density û . The second Hohenberg-Kohn theorem (HK-
II) states that the functional � � û������CüGý � û���� EüGý � û���� , � being the Hamiltonian,
will have the exact ground state energy ����������ý as a lower bound:� � û������ ���������ý (HK-II) � (2.111)
HK-II is equivalent to the variational principle. The so-called Kohn-Sham
Hamiltonian �
, which is applied in virtually all DFT applications, is just a sum
of one-electron operators (without electron-electron interaction):� � �"!$#&%�')(*,+.-0/�132�4 ú � /�1526 � (2.112)
Theory 48
Using ansatz (2.112) assumes that noninteracting electrons move in an external
local potential, called 798 , which has the property that its wave function - a single
Slater determinant of the lowest n (= number of electrons) orbitals - yields exactly
the same electron density as the exact interacting electron system with potential7 . This is correct, since the HK-I theorem states that 798 must be unique and this is
clearly independent of the form of the two electron interaction - totally neglected
in (2.112). The Kohn-Sham one-electron equations are::�;)<=,> ?A@�B5CED 7F8 @�B3CGIHKJMLN OQP N HRJMLNTS (2.113)
It is not clear if the Kohn-Sham orbitals H JMLN have any other physical significance
than the property that the sum of their squares adds up to the exact electron density.
Their orbital energies can not be related to the ionization energies like the Hartree-
Fock orbital energies (previous subsection) - except for the energy of the highest
occupied orbital which equals the negative of the exact ionization energy. It is
therefore hazardous to compare DFT and Hartree-Fock calculations at molecular
orbital level.
For details about the DFT method the reader is referred to refs [102, 103, 104].
The Time Dependent DFT (TD-DFT) method which can be applied to calculate
excited states is based on ”the fact that frequency dependent linear response of a
finite system with respect to a time-dependent perturbation has discrete poles at the
exact, correlated excitation energies of the unperturbed system.” [103] The mean
polarizibility U @�VWC is frequency dependent and describes the response of the dipole
moment to a time-dependent electric field with frequency V . Its relation to the elec-
tronic excitation energies V N OYX�Z�[N ; X�Z\[] and corresponding oscillator strength ^ Nis given by: U @�VWC O"_ N ^ NV ?N ; V ? S (2.114)
It can be seen from equation (2.114) that the mean polarizibility U @�VWC has poles atV O V N (= the excitation energy). In the Kohn-Sham formalism the exact linear
response to the time-dependent perturbation with frequency V�@�`C is expressed as
the linear density response. The ordinary Kohn-Sham orbitals (2.113) obtained in
a regular ground state calculation are involved. Their energy differences are shif-
ted towards the excitation energies (the poles in equation (2.114)) by a systematic
change in the perturbation frequency V . ”Hence, excitation energies are expressed
in terms of ground state properties and the problem whether density functional the-
ory can be applied to excited states is most elegantly circumvented.” [103] As long
as only low-lying valence states (not Rydberg states) are involved, the error of the
TD-DFT method is within a few tens of aAb . Therefore, the TD-DFT approach may
Theory 49
rival more sophisticated and much more costly wave function based approaches
(like CASPT2). Yet, comparison between these two approaches is not often in the
literature and the real performance of TD-DFT is difficult to judge.
The TD-DFT method has been applied in this work for the calculation of ex-
citation energies to compare them with the corresponding MS-CASPT2 excitation
spectrum. Furthermore, CASSCF optimized ground state geometries have been
compared with DFT optimizations. The applied particular functionals and basis
sets applied in this work will be described in chapter 3.
2.9 Calculation of the kinetic coupling terms T c�dfeand T c�ghe
This section deals with the numerical computation of the kinetic coupling terms
T ikjml and T ion�l using the CI and MO coefficients of a multiconfigurational wae func-
tion. In section 2.9.1 some general properties of T ikjml and T ipn�l are reviewed. In
section 2.9.2 our method of calculating T iqjml is presented and compared with dif-
ferent approaches found in literature. Finally, in section 2.9.3 it is described how
T ion�l can be calculated using T iqjml .2.9.1 General Properties of T r�sut and T rwvFtThe one-dimensional Schrodinger equation (2.21) with coupling elements defined
by equations (2.22) and (2.23) is considered. It is assumed that the electronic wave
functions xzyW{0| are real and orthonormal} y�~�x�yW{u|�����~�{�� (2.115)
If � is equal to � it follows:������ } y�~�x�y�~�|�� } y�~�x ������ y�~�|�� } ������ y�~�xzy�~�|��Y��� } y�~�x ������ y�~�|��Q�� } y�~�x ������ y�~�|����f� (2.116)
In the case that � is not equal to � the result is:������ } y�~$xzyW{0|�� } y�~�x ������ yW{A|�� } ��h��� y�~�x�yW{0|�� } y�~hx ������ yW{0|�� } yW{fx ������ y�~�|�W�Y�(2.117)
Since the x�yW{0| are real, � } y�~�x ������ yW{0|���� } yW{�x ������ y�~�|�� (2.118)
Theory 50
This means that the matrix � �k �¡ is antisymmetric with diagonal elements equal to
zero. According to Baer [52], � �£¢¤¡ can be written as
� �o¢¤¡�¥ ¦¦h§�¨ � �k �¡© ª�« ¬¨��®p¯w°f±�²�³K³K´�®pµ\¯p¶· � �q �¡�¸u� �k �¡© ª�« ¬±�²³�³K´�®pµ5¯p¶T¹ (2.119)
This means that � �o¢¤¡ is neither symmetric nor anti-symmetric. Equation (2.119) is
derived in section 2.9.3.
2.9.2 Calculation of the kinetic coupling term T º�»0¼ using a mul-
ticonfigurational wave function
The electronic multiconfigurational wave function ½�¾W¿uÀ can be expressed as a lin-
ear combination of state configurations ½zÁÃÂ�À½z¾W¿�Ä §�¨�Å À�¥"Æ ÂÈÇ ¿3ÂFÄ §�¨�Å ½zÁÃÂFÄ §�¨�Å À�É (2.120)
where the Ç ¿3Â are the CI coefficients of the Ê -th electronic state.
Substituting equation (2.120) in equation (2.22) leads to [62]:� �k �¡Ë ¿ Ä §�¨�Å ¥ÍÌ Æ Â Ç Ë ÂFÄ §�¨�Å ÁÃÂFÄ §�¨�Å ½ ¦¦�§�¨ ÆÏÎ Ç ¿ Î Ä §�¨uÅ Á Î Ä §�¨uÅ À¥ Æ Â Æ�Î Ç Ë ÂÑÐ ¦¦�§�¨ Ç ¿ ÎuÒ Ì\ÁÃÂÓÄ §�¨uÅ ½zÁ Î Ä §�¨�Å À© ª�« ¬ÔkÕ\Ö · Æ Â Æ�Î Ç Ë ÂFÄ §�¨�Å Ç ¿ Î Ä §�¨�Å Ì\ÁÃÂÓÄ §�¨uÅ ½ ¦¦�§�¨ Á Î Ä §�¨�Å À¥�Æ Â×Ç Ë Â Ð ¦¦�§�¨ Ç ¿5 ҩ ª�« ¬ØhÙoÚqÛÜmÝ
· Æ Â Æ�Î Ç Ë ÂFÄ §�¨�Å Ç ¿ Î Ä §�¨�Å Ì\ÁÃÂAÄ §�¨uÅ ½Þ¦¦�§�¨ Á Î Ä §�¨uÅ À© ª�« ¬ß Õ\Ö© ª�« ¬à ÙoÚqÛÜÝ ¹ (2.121)
In the last expression á �q �¡Ë ¿ is referred to as the CI term involving differentiation of
the CI coefficients and the term â �q �¡Ë ¿ contains derivatives of configurations or de-
terminants [62]. As will we shown in section 3.6.1, the â �k �¡Ë ¿ term ultimately leads
to integrals of the form Ì5ã�ä3½Kåå�æ�ç ½�ã ³ À where ã�ä and ã ³ are the orbitals by which the
determinants ÁÃÂ and Á Î differ and is, therefore, referred to as the MO term. Mat-
rix elements of determinants differing in more than two orbitals will vanish for the
following reason: Since åå�æ ç is a one-particle operator, according to the Condon-
Slater rules [88], the matrix elementsÌ\ÁÃÂFÄ §�¨�Å ½ ¦¦h§�¨ Á Î Ä §�¨�Å À
Theory 51
are non-zero only if the determinants èÃé and èëê differ in less than two orbitals.
Furthermore, the determinants èÃé are assumed to be real and orthonormal:ì èÃé�ízèëê�î�ïQðméêFñ (2.122)
This leads to equations similar to the ones reported in (2.116) and (2.118) of the
previous section: ì èòé�íÞóó�ô�õ èÃé�î�ï�ö�÷ (2.123)
ì èÃé�í óó�ô�õ èëê0î�ïùø ì èëêfí óó�ô�õ èÃé�î�ñ (2.124)
Whether the CI, úòûkü�ýþ�ÿ , or the MO term, �.ûqü�ýþ�ÿ , is the dominant one is rather arbit-
rary. For instance, ”diabatic” CASSCF orbitals [105, 106] change as little as pos-
sible as a function of geometry. Using the invariance of the CASSCF and MRCI
energies with respect to unitary transformations, they are generated by maximiz-
ing the overlap of CASSCF orbitals at a displaced geometry with the orbitals at
the reference geometry. Consequently, the relative contributions of the orbital and
CI contributions to the matrix elements of T û � ý are modified: The orbital contribu-
tion is minimized, and to a very good approximation the matrix elements of T û � ýcould be obtained from the CI-vectors alone. Therefore, given this smooth set of
”CASSCF” orbitals, rapid variations of the total wave function are confined to the
CI coefficients and can then be eliminated for instance by block-diagonalization
yielding quasi-diabatic states and energies for CASSCF and MRCI wavefunctions.
This ”direct” diabatization scheme has been applied e.g. to the photodissociation
of ozone [105, 106, 107] and H � S [108].
2.9.3 Calculation of the kinetic coupling term T�����
using T���
Substituting equation (2.120) in equation (2.23) yields� û � ýþ�ÿ � ô�õ � ï�� é�� þ é�� ó �óhô �õ � ÿ é������ é � ê � þ é � ô�õ � � ÿ ê � ô�õ ��ì èÃé � ô�õ � í ó �ó�ô �õ èëê � ô�õ � î������� é � ê � þ é � ô�õ � � óó�ô�õ � ÿ ê � ô�õ � � ì èòé � ô�õ � í óó�ô�õ èëê � ô�õ � î�ñ (2.125)
Instead of evaluating the last equation the following expression taken from ref.
[52] and derived below was used in this work to calculate� û � ý using
� ûkü�ý :� û � ý ï óó�ô�õ � ûkü�ý � � ûkü�ý � � ûqü�ý ñ (2.126)
Theory 52
Derivation of equation (2.126):Equation (2.126) can be proved by looking at a matrix element of � "!$# :%'&)(+*-, !,/. !0 &2143657 ,,8. 0 %9&)(/* ,,8. 0 &213;:=<?>�%9&)(@*A,,8. 0 & > 3B%9& > * ,,8. 0 &2143DC (2.127)
From (2.118) follows %9&)(/* ,,/. 0 & > 3 7FE %G,,8. 0 &)(@*H& > 3ICInserting the last expression in (2.127) leads to%9&)(+* , !,8. !0 &2143 EKJ %9&)(@* , !,8. !0 &213L:�% ,,8. 0 &)(+* ,,8. 0 &2143NM E < > E % ,,8. 0 &)(@*H& > 3B%'& > * ,,8. 0 &213O PRQ STVU W?XIY[Z\W]XU ^+_7 %9&)(@* , !,8. !0 &2143 E %9&)(/* , !,8. !0 &213 E % ,,8. 0 &)(@* ,,8. 0 &2143;:�% ,,8. 0 &)(+* ,,/. 0 &21B3�`7ba q. e. d.